Limited-Angle Tomography Reconstruction via Projector Guided 3D Diffusion
Researchers have developed TEMDiff, a new 3D diffusion-based framework for reconstructing 3D shapes from limited-angle electron tomography data. This method addresses the challenge of obtaining large, high-quality training datasets by utilizing simulated data that maps to Transmission Electron Microscopy (TEM) tilt series. TEMDiff demonstrates superior reconstruction quality compared to existing methods on simulated datasets and shows strong generalization to real-world TEM data, even with very narrow tilt ranges. AI
IMPACT Introduces a novel diffusion-based approach for 3D reconstruction in electron tomography, potentially improving material science and biological imaging.